AKAP7 / AKAP15 / AKAP18 (O43687) — review notes

2026-10-03. PAINT no-IBA backlog. Provider: affinage (gates clear, strong recall this time).

Scope first: this entry is the short isoforms only

O43687 holds AKAP18α (AKAP15) and AKAP18β — the short, N-terminally lipid-anchored,
plasma-membrane forms. The long cytosolic γ/δ forms are a separate entry, Q9P0M2:

[file:genes/human/AKAP7/AKAP7-uniprot.txt "IsoId=Q9P0M2-1; Sequence=External;"]

That matters because much of the affinage narrative — the 2H phosphoesterase that degrades 2-5A
(RNase L antiviral function), and the USP4/SERCA2 work in heart — belongs to the long isoforms.
None of it is imported here.

Outcome: 15 rows + 1 NEW → 5 MODIFY, 5 ACCEPT, 4 non-core, 1 UNDECIDED, 1 NEW

Refinements (MODIFY)

• GO:0051018 PKA binding (IDA, TAS, IEA) → GO:0034237 RII-subunit binding. The cleanest
  comparator result of the campaign: AKAP1, AKAP5 and AKAP9 all carry GO:0034237 (including IBA),
  and so does mouse Akap7, by IDA. Human AKAP7 was the outlier. Structural support is on this
  entry's own β isoform: PMID:27102985
• GO:0008104 → GO:0072659 protein localization to plasma membrane — the paper names the
  destination; GO:0072659 verified as a descendant of GO:0008104 via OLS.
• GO:0006811 monoatomic ion transport (TAS) → GO:0051924 regulation of calcium ion
  transport. Participation test: an AKAP moves no ions; it regulates the channel that does.

The NEW row

GO:0060090 molecular adaptor activity (IPI). AKAP15 makes both contacts itself — RII on one
side, the CaV1 C-terminal leucine zipper on the other:
PMID:14569017
Comparator: AKAP5 and AKAP9 carry GO:0060090 by IBA among others, so PAINT places it on the AKAP
lineage. Promoted to core_functions.molecular_function.

Kept but non-core: the IKs rows

PMID:11299204 gives four IDAs (K⁺ transport regulation, membrane repolarization, response to cAMP,
action potential). The result is real, but read the framing:
PMID:11299204
— three unrelated AKAPs, interchangeable, in heterologous cells, with a modest effect. That shows
AKAP15/18 can confer cAMP regulation when supplied, which is a property of being an AKAP, not that
it is the IKs anchor in myocytes. Kept, demoted to non-core.

GO:0001508 action potential → UNDECIDED. The abstract reports voltage-clamp currents and never
mentions action potentials. The full text might — publisher returns 403, no PMC copy — so per the
action definition it is UNDECIDED, not REMOVE.

A genuine conflict, recorded not resolved

• PMID:14569017
• PMID:23035250

The knockout deleted all isoforms. Both results stand; the open question is whether the peptides
hit other AKAPs or other AKAPs compensate in the knockout. Recorded as a BIOLOGY knowledge gap.

Recall

Unusually, affinage outperformed GOA: it surfaced the CaV1.1 leucine-zipper paper (PMID:11733497),
the knockout negative, and the mossy-fiber LTP work — none cited by any GO annotation. My own PubMed
search added PMID:14569017 (CaV1.2), which became the anchor for the NEW row.

The mossy-fiber result is curated on mouse (GO:0050804, GO:0098686, IDA+IMP) but not human —
recorded as a CURATION gap, consistent with how I treated the analogous AKAIN1 case.

Process

• Quotes checked whitespace-collapsed but otherwise character-exact: 24 distinct, 0 failures.
• GO_REF entries copied from the seeder.
• PR #3932's review bot failed on its usage quota ("session limit, resets 8pm UTC"), not on the
  PR; re-run after the reset.

Round 2 (review feedback)

All three IMPORTANT items were right.

1. GO:0060090 was the adaptor root. Both clients are proteins in one cAMP-PKA pathway, so
   GO:0035591 signaling adaptor activity (GO:0060090 > GO:0030674 > GO:0035591), matching
   the AKAP12 sibling. Added qualifier: enables and the IPI partner — rat CaV1.2 (P22002,
   verified via API), because that is the channel PMID:14569017 used. A human accession would have
   misstated what was tested.
2. Three cited records were never committed (9545239, 10613906, 11299204) — I read them locally
   and never git added them. Worse, the same gap reached main via AKAIN1 (PMID:25653177);
   fixed separately in #3939. Cause: my guard checked only staged ⊆ cited. It now also checks
   cited ⊆ (HEAD ∪ staged). Its first fix used origin/main and wrongly flagged records committed
   earlier on this branch; corrected to HEAD.
3. core_functions had dropped GO:0034237 when the adaptor took the single MF slot, despite
   three MODIFYs to it on crystallographic evidence. Added as a second core function.

Suggestions taken: ion-transport replacement → GO:1901385 regulation of voltage-gated calcium
channel activity (names the potentiation mechanism); description no longer describes database
structure. On GO:0071320, kept as non-core but answered the participation point explicitly — the
cAMP response is PKA phosphorylating the channel, and the AKAP supplies the positioning that step
depends on (the scaffold case), whereas for ion transport the AKAP does none of the work.